refactor example/c/lib_buffer_unpack.c

[n1tehawk]

The example has some duplicated code that somewhat distracts from
the main processing loop. I think placing this into a separate
function improves readability of the code.

[n1tehawk]

CI tests indicate that the previous commit made the "buf" variable obsolete (unused), potentially causing warnings.
n1tehawk@871a796 fixes that

[nobu-k]

assert macro can be disabled and this line could completely be removed. So, this has to be like the original code:

bool expanded = msgpack_unpacker_reserve_buffer(unpacker, request_size);
assert(expanded);

[nobu-k]

and it generates unused variable error (-Werror) :) umm. I'll talk to @redboltz about this.

[nobu-k]

msgpack_unpacker_reserve_buffer(unpacker, request_size)
    || (assert(0 && "msgpack_unpacker_reserve_buffer failed"), 0);

@redboltz Is this acceptable to you? Do you have a better idea?

[n1tehawk]

Oh... But that's a problem that also affects the original code, right?

[n1tehawk]

Since msgpack-c already has a MSGPACK_UNUSED macro in https://github.com/msgpack/msgpack-c/blob/master/include/msgpack/util.h#L21, it might be a saner approch to simply mark the local variable as "potentially unused":

bool expanded = msgpack_unpacker_reserve_buffer(unpacker, request_size);
assert(expanded);
MSGPACK_UNUSED(expanded);

EDIT:
A "clean" way to solve this is command-query separation. In this particular case that's easy to achieve:

if (msgpack_unpacker_buffer_capacity(unpacker) < request_size) {
    msgpack_unpacker_reserve_buffer(unpacker, request_size);
    assert(msgpack_unpacker_buffer_capacity(unpacker) >= request_size);
}

[nobu-k]

Your code looks much nicer than the previous one. Thanks! I'll merge this after fixing the problem above.

[n1tehawk]

Thanks! :) And nice catch with the assert() evaluation.

We might even further reduce the reading loop to a single call of receiver_to_unpacker(), like this;

while (true) {
    recv_len = receiver_to_unpacker(r, EACH_RECV_SIZE, unp);
    if (recv_len == 0) break; // (reached end of input)

Regards, NiteHawk

[nobu-k]

We might even further reduce the reading loop to a single call of receiver_to_unpacker(), like this;

I think it seems better. I'll apply the change when I merge this PR.

[n1tehawk]

I've incorporated the suggested changes into another commit. The only remaining issue that needs to be addressed is that assert() logic.

[nobu-k]

Thank you!

[n1tehawk]

One more thing: I'm wondering if unpack() (in lib_buffer_unpack.c) is missing a msgpack_unpacker_free(unp); at the end?

[redboltz]

One more thing: I'm wondering if unpack() (in lib_buffer_unpack.c) is missing a msgpack_unpacker_free(unp); at the end?

@n1tehawk, thank you for pointing out the problem. You are right. I checked the executable using valgrind, then resource leaks are detected.

[n1tehawk]

Hi @redboltz!

I'll prepare another commit then - which will contain the free() call, and also addresses the assert() stuff discussed before.

[redboltz]

Merged. Thanks you @n1tehawk :)